Académique Documents
Professionnel Documents
Culture Documents
Application
Abdo Salah,
Nishant Singh,
Rajen Paudel,
Students,Master of Science in Communication Engineering,
Technical University of Munich,
Munich
May 5, 2014
Contents
1 Concept and Theories
2 Spectrum Analysis
Chapter 1
Concept and Theories
1. What is the function of a Modulator?What is the difference between a
ADC and a Nyquist Rate ADC? What is the best application for a ADC?
Why?
A Modulator converts analog signals into a low resolution,noise shaped,digital signal.A ADC is classified as a Oversampling ADCs , wherein input signal is sampled at
a much higher frequency than Nyquist Frequency, usually the sampling frequency can be
128 to 512 times higher than that of Nyquist Frequency, the ratio of sampling frequency
to the Nyquist Frequency is termed as OverSampling Ratio or OSR, the required OSR is
decided on the application requirements.
Nyquist ADCs are the one who sample input signals at ,or close to (but higher),Nyquist
Frequency.
ADCs are mostly used in industrial application,or in voiceband-audio , wherein a a
high resolution in data is required (>15 bits) but the frequency requirement is low (less
than 100KHz).
ADC
Sampling Rate is higher than Nyquist Frequency (of the order of 100 times above the
input frequency)
Is ideal for low frequency signals, as it offers
a higher resolution (upto 20 bits) with simple
cicuit implementation
The linear model is accurate if the dependency of the error e[n] of the input signal is taken
into consideration. Under the common assumption of e[n] being white noise independent
3
Chapter 2
Spectrum Analysis
1. Give the equation that defines DFT.
DFT or Discrete Fourier Transform of a signal can be defined as ,
X(k )
PN 1
N
1
X
x(tn )ejk tn ,
k = 0, 1, 2, . . . , N 1, , where,
n=0
2. DFT has to be used to analyze the frequency spectrum of the output bit
stream. Matlab provides a FFT function instead of DFT. What is the relationship between these two?
There is no difference in concepts between FFT and DFT. FFT is exactly DFT but
because DFT is not an efficient way to determine the frequency content of a time domain
sequence when the number of points exceed some limits such as thousand or hundreds
points, an algorithm called FFT (fast Fourier transform ) was introduced. There are many
algorithms to calculate the FFT but the most famous on is radix-2 FFT . The radix-2
FFT algorithm is a very efficient process for performing DFTs under the constraint that
the DFT size be an integral power of two. (That is, the number of points in the transform
is N = 2k , where k is some positive integer.)
While DFT can also be applied to any complex valued series, in practice for large series
it can take considerable time to compute. The time taken being proportional to square of
number of elements in the series. So the faster algorithm is desired in most applications,
which is FFT , and the computing time is proportional to Nlog(2n).
3. Explain the term window function. Why do we need it?How does it affect the
result?
Window function is a mathematical function that is of 0 value outside the interval. It
is observed that in aperiodic signals the FFT has signal much more dispersed.The dis6
persed nature of FFT makes it difficult to identify the frequency content of the input
signal.Since in practical implementations most of the signals are aperiodic within the
pre-defined datablock, the FFT would suffer from the dispersal of signal enrgy, which is
termed as leakage.
A window is applied for correction of leakage. A window function is shaped such that it is
exactly zero before and after the data block, and has some sort of signal shaping function
in between.This function is multiplied by time data block which forces the signal to be
periodic.Thus as the signal is now periodic, the FFT of it does not show any leakage.
However the FFT needs to multiplied by a weighing factor to correct the amplitude levels
, as the original amplitude levels of the input signal gets distorted because of the window
function, so after FFT , an additional multiplication by a weighing factor is needed to
restore original signal levels. The figure below depicts the application of Hanning window,
the change in FFT leakage is noticeable.
4. What is SNR? Explain the way to calculate SNR from a frequency spectrum
resulted from DFT with window function for a modulator with sinusoidal
input signa and implement this as a Matlab Script.
SNR stands for Signal to noise ratio, of a given signal, it is given by following formula,
SN R = Signalpower
N oisepower
7
The SNR can be estimated from DFT by computing ratio of power in the signal bins to
power in the in-band noise bins.For a high SNR number of bins occupied by signal should
be a relatively small (less than 20% number of in-band noise bins). When using Hanning window, the signal would occupy 3 bins, so we should compute power in all three
bins and take mean of that. By implementing the script (given at the end) , it was
seen that windowing can increase SNR significantly, for a sampling frequency of 4000
Hz, and input frequency of 543 Hz, a window of length 256 was used (corresponding to
approximately 64.OSR). It was seen that without windowing a SNR of 4.5 was achieved,
on windowing the same signal , the SNR achieved was 27.4 . The SNR can further be
increased by increasing the window length. The figure belows show the input waveform
, windowed waveform, FFT without windowing, and FFT with windowing waveforms,
The Matlab script used for getting above result is given as under,
clear all;
Fs = 4000;
fin=543;
T=1/Fs;
8